@hasna/knowledge 0.2.15 → 0.2.17

package/README.md

@@ -97,6 +97,11 @@ open-knowledge embeddings search "company wiki policy" --scope project --json
 # Hybrid search over source chunks, generated wiki pages, indexes, and optional vectors
 open-knowledge search "company wiki policy" --scope project --json
 open-knowledge search "company wiki policy" --scope project --semantic --json
+open-knowledge search "company wiki policy" --scope project --context --json
+
+# Build a citation answer/context draft for a prompt
+open-knowledge ask "How do we cite handbook policy?" --scope project --json
+knowledge "How do we cite handbook policy?" --scope project --json
 ```
 
 ## Commands
@@ -245,6 +250,7 @@ permission/path/delete metadata, and records a local run ledger.
 ```bash
 open-knowledge search <query> [--scope project] [--limit <n>] [--json]
 open-knowledge search <query> --semantic [--model openai:text-embedding-3-small] [--scope project] [--json]
+open-knowledge search <query> --context [--semantic] [--scope project] [--json]
 ```
 Run hybrid search over `chunks_fts`, generated wiki chunks, wiki/index catalog
 rows, and optional vector results. The default path is local-only keyword and
@@ -252,6 +258,24 @@ catalog search. `--semantic` embeds the query and merges vector results from
 `vector_index_entries`, preserving source refs, artifact URIs, citations,
 revision/hash metadata, and provenance in each structured result.
 
+`--context` returns a reranked context pack for agents: selected excerpts,
+assembled citations, freshness and permission notes, graph evidence from
+`citations`/`wiki_backlinks`, and final rerank scores. This is the shape future
+`knowledge <prompt>` flows should send to a model instead of raw search rows.
+
+### ask / build
+```bash
+open-knowledge ask <prompt> [--scope project] [--json]
+open-knowledge build <prompt> [--generate] [--model default|provider:model] [--scope project] [--json]
+knowledge <prompt> [--scope project] [--json]
+```
+Build an agent-native prompt run. The command first creates a read-only context
+pack, returns a local citation draft by default, records a run ledger in
+`runs`/`run_events`, and proposes durable wiki updates without writing them.
+`--generate` explicitly calls AI SDK text generation; `--fake --generate` keeps
+the flow deterministic for local tests. `--approve-write` records approval
+intent, but durable wiki writes remain deferred to the wiki compile/write task.
+
 ### safety
 ```bash
 open-knowledge safety status [--scope project] [--json]
@@ -331,7 +355,10 @@ The MCP server exposes item tools (`ok_add`, `ok_list`, `ok_get`, `ok_update`,
 `ok_storage_status`), provider/embedding tools (`ok_provider_status`,
 `ok_provider_models`, `ok_embeddings_status`, `ok_embeddings_index`,
 `ok_semantic_search`), hybrid retrieval (`ok_search`), and source-ref
-parsing/resolution (`ok_parse_source_ref`, `ok_resolve_source`).
+parsing/resolution (`ok_parse_source_ref`, `ok_resolve_source`). The
+`knowledge_search` MCP tool returns reranked citation context packs for agent
+prompts, and `knowledge_ask` runs the same prompt flow exposed by
+`open-knowledge ask`.
 
 ## Source And Artifact Boundary